linux-kernel - Re: [PATCH 1/3] hwmon: (sht3x)remove sht3x_platform

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <54ca1f3c-4d24-96f9-271e-30f97f5b75f4@roeck-us.net>
Date:   Thu, 15 Jun 2023 07:56:19 -0700
From:   Guenter Roeck <linux@...ck-us.net>
To:     JuenKit Yip <JuenKit_Yip@...mail.com>
Cc:     linux-hwmon@...r.kernel.org, linux-kernel@...r.kernel.org,
        jdelvare@...e.com
Subject: Re: [PATCH 1/3] hwmon: (sht3x)remove sht3x_platform_data

On 6/14/23 16:54, JuenKit Yip wrote:
> 
> 在 2023/6/15 3:15, Guenter Roeck 写道:
>> On 6/14/23 08:02, JuenKit Yip wrote:
>>>
>>> 在 2023/6/14 20:57, Guenter Roeck 写道:
>>>> On 6/13/23 23:24, JuenKit Yip wrote:
>>>>> Since no in-tree driver supports it, the sht3x_platform_data was
>>>>> removed.
>>>>>
>>>>> - "blocking_io" property and its related code have been removed, and
>>>>> Single-Shot mode should be blocking in default.
>>>>>
>>>>> - "high-precision" property has been replaced to "repeatability" for
>>>>> matching datasheet.
>>>>>
>>>>
>>>> That needs to be three patches.
>>>
>>> Patch 1: remove sht3x_platform_data and its header file
>>>
>>> Patch 2: move "blocking_io" to struct sht3x_data
>>>
>> Essentially merge it with update_interval==0 since (if I understand
>> correctly) blocking mode and update_interval==0 will be equivalent.
>> With that in mind, a separate "blocking_io" variable should no
>> longer be needed.
>>
> I reviewed the datasheet again, update_interval == 0 means Single-Shot
> 
> mode which owns blocking(clock strench) and non-blocking(non-clock strench)
> 
> options. If master supports clock-strench( I don't know how to detect it),
> 
> the property may be reserved.
> 

I see.

In practice, blocking mode was never really used, at least not in the
upstream kernel, since platform data was not set from any in-kernel
code. Given that it is pretty much untested, I would suggest to always
use non-blocking mode. This is only relevant if the chip is in
single-shot mode, so worst case the user would have to wait a bit longer
for results in that mode. I find that acceptable over the risk involved
when trying to use/support blocking mode.

Guenter